The "norm" parameter of TfidfVectorizer in scikit-learn is used to specify the normalization method applied to the resulting TF-IDF vectors. It controls whether the vectors should be normalized to have unit norm (L2 normalization) or left unnormalized (None).
- 'l2': The sum of squares of vector elements is 1. The cosine similarity between two vectors is their dot product when the L2 norm has been applied.
- 'l1': The sum of the absolute values of vector elements is 1.